e7200 @ 3.8ghz -P5QL Pro (for 24/7 use)

[AiponGkooja]

I have it at 400mhz x 9.5 = 3.8ghz

1.32vcore
1.4v fsb termination
1.4v northbridge
1.5v southbridge
1.8v ram
5-4-4-12 ram 1:1 800mhz

Prime95 (20 minutes so far)
CoreTemp 58C (although it went to 63 briefly while I was typing this...)
RealTemp 53/63C (95/105 tjMax) (59/68 briefly while I was typing this...)

Idle Temp ~37C



Anything else I should test? Or just keep prime running for a while? Are temps ok? Voltages? From what I've read so far, it all seems within "~reasonable" limits. Will it e good for 24/7 use though (assuming it keeps chugging through prime for a while)?

 

[Concillian]

If Prime goes for a good while, like 24hrs you should be good.

Personally for a 24 / 7 / 365 OC, I push a specific voltage to find the bleeding edge of stability, then back off 5-10 MHz on the FSB from there to account for warm days, eventual dust buildup, etc...

Though I seem to be a lot less extreme than most here. I like to consider it practical, as you won't really notice that extra squeeze, but you will notice when that extra squeeze fails a few months from now due a heat wave. It always seems to happen when you have some appointment to keep and don't want to mess with troubleshooting the computer.

 

[VirtualLarry]

I second the idea of pushing it to the extreme, and then backing off a bit to give you some headroom.

I backed off 5 FSB on my E4400, I could prime95 at 2.85, but I backed down to 2.8 for stability. Same with my E2140 @ 3.2, I had them at 3.28 for a bit, but backed down.

Still, about monthly, I have issues with my E2140. Sometimes it spontainously reboots, sometimes it bluescreens. The E4400 doesn't do either, it just keeps running. I don't know what it is, I can stresstest for multiple 24hr periods and it's fine. Just that the E2140s don't seem to like uptimes of greater than a month.

I did have problems with almost daily reboots, until I replaced the UPS that it was plugged into. Something strange with the UPS. I replaced it with the same model, so it's possible that the UPS is still giving problems. It's not overloaded.

 

[AiponGkooja]

After that first post, it had an error at about 32 minutes in prime95. I upped the vcore to 1.33 and then 1.336, but for some reason the tests last progressively LESS time... 11 minutes and 6 minutes. Am I missing something?


Edit: Sorry, lied about my SB, it's 1.5v. I put in 1.2v, but didn't notice the min was 1.5v... whoops.

 

[myocardia]

Originally posted by: AiponGkooja
After that first post, it had an error at about 32 minutes in prime95. I upped the vcore to 1.33 and then 1.336, but for some reason the tests last progressively LESS time... 11 minutes and 6 minutes. Am I missing something?

That usually means that you're overheating your northbridge or VRMs, or both. There are alot of coincidences in overclocking, though, so it doesn't necessarily mean that either of those two are your problem. That would be the first things I would check, though.

 

[AiponGkooja]

Yeah, I moved my NB down to 1.3V, since 1.4 was max, and that seemed to help.

I honestly don't know what VRM is or how to check if they're overheating though... I will keep searching however. And thank you for the suggestion.

 

[JPB]

This makes me feel ok about overclocking mine. If the damn thing ever gets here

 

[error8]

I would personally use Linpack to test for stability, instead of the 24 hours Prime 95 marathon. I can't keep my computer occupied such an extended period of time, so Linpack tells me much faster if I'm stable or not ( aournd 15 minutes, 1/2 memory, 10 tests ). Passing at least 5 out of 10 test, can mean that the cpu is totally stable, in any situation except Linpack, of course. Another good stability test is OCCT. Try it out.

 

[AiponGkooja]

Thank you for the tip. OCCT is working much better for me, mostly because with prime95 I would get impatient and end it after 20 minutes, haha. With OCCT I know it has to run for an hour, so I just plan accordingly.

I'm still not sure what voltages are safe for my setup. And where can I check temperatures of the things that fsb termination voltage and northbridge voltage affect? I have them at 1.3V each now, but I can't tell anything except whether or not it's helping my system stability. I just do not want to sacrifice part life-time for stability.

 

[PlasmaBomb]

Intel burn test is a handy way of doing linpack testing if you need one...

The test that gets things hottest (at least for me) is Coredamage.

 

[Comdrpopnfresh]

DO NOT RUN TAT ON CORE2 CHIPS! Check intel whitepapers- they make a point of saying it is a big no-no. Also, use speedfan or realtemp, as they measure temps based off of tcase- which is what thermal specs for core2 chips are rated off of.... hwmonitor (i think) and coretemp use Tjunction, and will not be as accurate- you can offset those, but the readings between tcase and tjunction cannot be calibrated- they will differ through a temperature change.

Your fsb termination @ 1.4V is high- that's threshold from what I've read. For an e7200 you shouldn't have to touch that much, or at all- that's more important on quad cores.

Obviously, make sure power saving features are off. I'd suggest dropping ur cpu multiplier to 6x and leaving the vcore where it is, maybe turn it up to 1.4 in bios (the droop should keep it safe). Then tune the fsb upwards, only touching nb voltage when you cannot boot. I found that my p5n-d has a very stable point @ 1700fsb, and I actually need to increase the nb voltage by .2V to run the processor at a slower speed @ fsb~1650. If you find one of those stable points, and know you're boards max fsb, you'll be much more prepared to isolate what components/voltages are making an attempted overclock unstable. If you find a stable fsb region that your e7200 cannot run at 9.5x multiplier on, just manually set it, leave speedstep disabled, and enable c1st when your sure of stability and ready to run 24/7. This'll let the processor run between 6x multi and whatever you manually set.
for example: I told you my board is very stable @ 1700 fsb, but my e7200 has bad thermals @ 9.5x w/ that fsb. So I set it to 9x, and I run @ 3825mhz under usage and 2.6-something ghz idle/low-load.
Here are my volts:
vcore: 1.3875V in bios, cpu-z reads 1.36V after droop.
vtt/fsb-term: 1.2V
NB: 1.2V
SB: 1.5V

@ full load my temps, reported by a calibrated speedfan are in the mid 50's, low 60's on a hot day.

 

[AiponGkooja]

What is TAT? I just run prime95 and OCCT now. As I've heard those a lot, so I trust them. I haven't tried the other 2 or 3 that people suggested because I've never heard of them.

Thank you very much for the detailed overclocking methods!

Yeah I was stuck on just leaving it at 400mhz and 9.5 and just getting that stable... never thought of making it higher to actually have a more stable oc... heh.

Sorry for the confusion, but is that 1700 the (cpu clock * 4)? So yours is more stable at 425?

Also, my RAM is rated at 800, 1.8v, so would that 425 mess anything up (not sure what ratio I'd have to run since I can't check my BIOS atm)? Or wouldn't I really notice a performance difference?

I will start over with minimum voltage and low multiplier and try the fsb boost again if you think that's the best idea at this point.

 

[AiponGkooja]

Short update.

400 x 9.5
1.33x vcore
1.3v fsb termination
1.2v northbridge
1.5v southbridge
1.8v ram
5-4-4-12 ram 1:1 800mhz

Ran OCCT - 1 hour test, stable.
Ran OCCT - infinite test, 4 hours, fail.
Upped vcore to 1.334x.
Ran OCCT - infinite test, less than a minute, fail.
Ran OCCT - infinite test, 2.5 hours, fail.

Suggestions?

Thank you.

Edit: Heat was 62C max during 4 hour OCCT run. (think it uses tjMax 100)

 

[PlasmaBomb]

TAT =Intel Thermal Analysis Tool

 

[nyker96]

I recommend LinX testing as well, from my experience, for instance, my E7200 is stable at 1.35v//3.6Ghz on Prime(8hr) yet can only be stable on Linx with 3.52Ghz. And LinX runs the temp much higher as well. This test has became the gold standard for my stability toolkit. I highly recommend that you try it.

 

[AiponGkooja]

How much higher does it run the temp? I just bumped the vcore up to 1.344 (cpu-z) and it has been stable through 8 hours of OCCT. Max temp at 62 (with tjMax of 100 I think...). I may try reseating my Xig 1283 (possibly lapping it a little first), and experimenting with the TIM application, because I kind of rushed through the installation the first time around... But I don't want to try something that is going to throw my temps to 70+. My goal is under 60, but I'm tolerant of 60-65 for full ridiculous torture testing load.

Also, what is the best way to check other temps on my board? I want to see if the NB or SB are getting too hot as well. Although I do have them at pretty low voltage now finally. Any other temps I should be concerned about?

 

[nyker96]

Originally posted by: AiponGkooja
How much higher does it run the temp? I just bumped the vcore up to 1.344 (cpu-z) and it has been stable through 8 hours of OCCT. Max temp at 62 (with tjMax of 100 I think...). I may try reseating my Xig 1283 (possibly lapping it a little first), and experimenting with the TIM application, because I kind of rushed through the installation the first time around... But I don't want to try something that is going to throw my temps to 70+. My goal is under 60, but I'm tolerant of 60-65 for full ridiculous torture testing load.

Also, what is the best way to check other temps on my board? I want to see if the NB or SB are getting too hot too. Although I do have them at pretty low voltage now finally. Any other temps I should be concerned about?

It runs about 5 degree higher for me at load. I use Vendetta 2 cooler which is similar to S1283. I always do OCCT/ortho 1st to make sure temp is ok then run LinX that always pushes the temp higher. But one word of moderation is that it is probably ok to not get your rig 24hr LinX stable to be perfectly stable on everyday apps. LinX is just a higher stability standard than OCCT/Ortho.

As for temp, I keep mine below 70C or so. BTW, Intel published TjMax for E7200 is 95C some program gets that one wrong. One thing I hate about DT coolers is the TIM application never seems to be easy. I almost have to double up the amount as for a regular cooler. You can use speedfan for NB/SB temperatures, they have profiles for most motherboards.

 

[AiponGkooja]

That's cool. I think I won't even bother with LinX then... I mean, I mostly just do gaming anyway. *Shrug*

Also, good news about tjmax then, as that means my real temps are closer to 57C load, so that makes me happy. Still might try reseating HS though, just for fun. Heh.

Thanks for all the info!

 

[AiponGkooja]

Short update.

400 x 9.5
1.344 vcore (cpu-z)
1.3v fsb termination (1.5 max)
1.2v northbridge (1.4 max)
1.5v southbridge (1.7 max)
1.8v ram
5-4-4-12 ram 1:1 800mhz

Ran OCCT - 1 hour test, stable.
Ran OCCT - infinite test, 8 hours, stable.
Ran Prime95 - 1.5 hours, fail
Upped vcore to 1.35x
Ran OCCT - infinite test, 1 hour, fail
Upped vcore to 1.360 (cpu-z)
Running OCCT, will post results later today


Any suggestions for if the test runs less time again? Should I drop back to 1.344 and run the tests one more time to make sure it really is the "most stable" vcore, and then fiddle with other voltages? If so which ones?

 

[AiponGkooja]

Another quick update...

400 x 9.5
1.36125 vcore BIOS (1.360 cpu-z)
1.3v fsb termination (1.5 max)
1.2v northbridge (1.4 max)
1.5v southbridge (1.7 max)
1.8v ram
5-4-4-12 ram 1:1 800mhz

Ran OCCT - infinite test, 12 hours, stable.
Running prime95 - 4 hours so far, stable (going to let it run over night at least)

Does it seem like this is probably the best I can do? Or based on my other results, should I be able to fiddle with other voltages and drop my cpu vcore? I don't really like that it's over 1.36, as that is Intel's tolerant "max".

I don't think heat will be an issue with anything else that I do at this point, since I shouldn't have to raise the cpu voltage any further. It hits about 57C MAX on full load (usually closer to 53-55), and I will probably still reseat my HS after I confirm stability.

 

[AiponGkooja]

Prime95 failed after 9 hours... Should I even care? heh.

 

[Concillian]

Only if you would ever be bothered if your computer froze or randomly rebooted while doing something important.

Prime95 is good and all, but I've had situations where a computer passed 24 hours Prime95 and I found a game that would randomly crash out, then backing 5FSB down fixed it.

This is why I said previously in the thread that once you find your stable spot, I advise backing off from there a bit.

 

[AiponGkooja]

Awww, but that means I have to make it stable at 3900 mhz, so I can back it down to 3800. hehe :-D

 

[AiponGkooja]

Finally stable I think. I did change the fsb to 405, and bumped the vcore up to 1.36875 in BIOS (bounces between 1.360-1.368 in cpu-z). prime95 stable for 23 hours. After that I just dropped it back down to 400, and left the vcore, and am running OCCT for 24 hours hopefully (then maybe prime95 again).

So right now:

400x9.5 (I just really like round numbers I guess...)
vcore - 1.36875V (not overly excited about that...)
temp - 58C max on full load
FSB termination - 1.3V (1.5V max in BIOS)
NB - 1.3V (1.4V max in BIOS)
SB - minimum 1.5V (1.7V max in BIOS)
Mem - 1.8V, 5-4-4-12, 1:1, 800mhz


Are those volts still unsafe? Or are they close enough that I probably don't have to worry about it? Should I continue to tweak the other voltages (p5ql pro is not very oc-friendly... FSB term & NB go in 0.1V increments, SB in 0.2V increment)? I think I'm going to get a couple tiny fans to ziptie to my chipset heatsinks just because.

Also, one more quick question, does temperature have much to do with stability? Or is it just volts? with temperature being a byproduct of increased volts, but not necessarily causing instability itself? I ask because my room was kind of cool yesterday/today when stability testing, so I don't know if, when it is warmer and temps hit maybe 62-63C instead of 58C, it would make be less stable...

 

[AiponGkooja]

OCCT stable for 24 hours. Running prime95 for another day now.

Any comments on my last post? I read 1.36 was max safe voltage... but then again I see people put over 1.4v for 24/7 setups, so I don't know what to believe. At the same time, I guess overclocking at all is somewhat hazardous since it voids the warranty, heh.

Thoughts?

Thank you all!


Edit:

Thought I'd put some temps *shrug*

HWMonitor Max temps under full load:
SYSTIN - 33C
CPUTIN - 63C
AUXTIN - 25C
Core 0 - 58C
Core 1 - 58C

Real Temp:
CPU - 57C (tjMax 95)